Conversation

Notices

1. Embed this notice
   on-lain ✔ᵛᵉʳᶦᶠᶦᵉᵈ (lain@lain.com)'s status on Monday, 09-Oct-2023 01:32:57 JST on-lain ✔ᵛᵉʳᶦᶠᶦᵉᵈ
   Remember that LLMs are all just prediction engines
   • Sexy Moon and ロミンちゃん like this.
   • Embed this notice
     kaia (kaia@brotka.st)'s status on Monday, 09-Oct-2023 01:34:20 JST kaia

     in reply to
     @lain I predict them going to jail
   • Embed this notice
     Hoss Delgado (hoss@shitpost.cloud)'s status on Monday, 09-Oct-2023 01:36:53 JST Hoss Delgado

     in reply to
     It'd be nice if they would just spare you the stupid fucking lecture and say "no".
     on-lain ✔ᵛᵉʳᶦᶠᶦᵉᵈ likes this.
   • Embed this notice
     on-lain ✔ᵛᵉʳᶦᶠᶦᵉᵈ (lain@lain.com)'s status on Monday, 09-Oct-2023 01:38:25 JST on-lain ✔ᵛᵉʳᶦᶠᶦᵉᵈ

     in reply to

     • waltercool
     @waltercool you can't just tell chatgpt to just continue predicting, so it's not possible to do what i did
   • Embed this notice
     waltercool (waltercool@pl.slash.cl)'s status on Monday, 09-Oct-2023 01:38:27 JST waltercool

     in reply to
     @lain Well, if you are using LLaMa2, Dan mode is pretty easy.
     
     ChatGPT it been trained to prevent Dan mode and enforce some censorship/bias on the information
   • Embed this notice
     Wilmhit until I woe (wilmhit@awawa.cat)'s status on Monday, 09-Oct-2023 02:04:09 JST Wilmhit until I woe

     in reply to
     @lain aren't we all?
     on-lain ✔ᵛᵉʳᶦᶠᶦᵉᵈ likes this.
   • Embed this notice
     guizzy (in exile) (guizzy@shitposter.club)'s status on Monday, 09-Oct-2023 02:05:38 JST guizzy (in exile)

     in reply to
     @lain Oobabooga in chat mode has a field for "Start reply with", you can tell it to start with "Sure thing, here's how you do it:". You can also add negative prompts or ban tokens (like "As an AI assistant")
     on-lain ✔ᵛᵉʳᶦᶠᶦᵉᵈ likes this.